Skip to contents

Obtain Kenward-Roger Adjustment Components

Source: R/kenwardroger.R
h_get_kr_comp.Rd

Obtains the components needed downstream for the computation of Kenward-Roger degrees of freedom. Used in mmrm() fitting if method is "Kenward-Roger".

Usage 

h_get_kr_comp(tmb_data, theta, linear = FALSE, w = NULL)

Arguments

tmb_data

(mmrm_tmb_data)
produced by h_mmrm_tmb_data().

theta

(numeric)
theta estimate.

linear

(flag)
whether to omit second derivatives and the R component.

w

(matrix or NULL)
covariance of the covariance parameters. Supply with linear = TRUE to contract Q without constructing its blocks.

Value

Named list with elements:

  • P: matrix of \(P\) component.

  • Q: matrix of \(Q\) component.

  • R: matrix of \(R\) component, or NULL when linear = TRUE.

  • S_Q: contracted Q sum when w is supplied; Q and R are then NULL.

Details

the function returns a named list, \(P\), \(Q\) and \(R\), which corresponds to the paper in 1997. The matrices are stacked in columns so that \(P\), \(Q\) and \(R\) has the same column number(number of beta parameters). The number of rows, is dependent on the total number of theta and number of groups, if the fit is a grouped mmrm. For \(P\) matrix, it is stacked sequentially. For \(Q\) and \(R\) matrix, it is stacked so that the \(Q_{ij}\) and \(R_{ij}\) is stacked from \(j\) then to \(i\), i.e. \(R_{i1}\), \(R_{i2}\), etc. \(Q\) and \(R\) only contains intra-group results and inter-group results should be all zero matrices so they are not stacked in the result.

Supplying w for linear KR instead retains P and contracts Q into the single matrix S_Q, without storing Q blocks, see the section "Contracted linear covariance adjustment" in vignette("kenward", package = "mmrm"). The symmetric part of w is used, and non-finite entries in w give a non-finite S_Q. mmrm() always supplies w for linear KR. Calling with linear = TRUE but without w gives the pairwise P and Q blocks, which tests use as an independent reference.